Algorithmic due process

Lens adapted to this topic: How existing due process can govern algorithmic decisions

The mainstream legal-policy approach treats algorithmic decisions as subject to familiar procedural requirements rather than as exempt from them. It emphasizes notice, reasons, an opportunity to be heard, assessment of error risks, meaningful human review, testing, and institutional accountability. This view accepts that automation may improve administration or reduce some forms of bias, but argues that those benefits do not displace constitutional and administrative-law safeguards.

• When government uses AI to deny benefits, bail, immigration status, or other protected interests, the due process analysis should still ask whether the person received adequate notice and an opportunity to be heard.
• Automated systems can undermine due process when people cannot meaningfully understand the basis of a decision or when decision-makers presume that computer outputs are infallible.
• Procedural protections can include interpretable and adequately tested systems, review of error risks, explanation, and mechanisms for correcting inaccurate outcomes.
• Algorithmic tools may help modernize legal institutions, allocate constrained resources, and potentially reduce bias, but their use requires safeguards and accountability.

Lens adapted to this topic: Claims that procedural safeguards may not fix automated adjudication

The alternative perspective argues that algorithmic due process should not be reduced to disclosure, accuracy, or formal opportunities to appeal. It focuses on how automated systems encode institutional priorities, reshape power, and turn context-dependent legal judgment into prediction or computation. From this view, democratic oversight and sociotechnical evaluation may be more important than making an opaque model merely more explainable.

• Fairness metrics and technical fixes may have only a tenuous relationship to people’s real-world outcomes; evaluation should examine the wider organizational and political system in which an algorithm operates.
• AI can reproduce or intensify injustice because it helps shape social and political institutions, not merely because an individual model contains a measurable bias.
• Law relies on purpose, moral commitment, language, and social context, which critics argue are structurally incompatible with treating legal reasoning as a fully computable process.
• Automated adjudication can diminish the legal subject by replacing individualized judgment with prediction, particularly in criminal-law contexts where risk forecasts influence decisions.
• Due process should remain inalienable rather than being weakened through contractual or technical arrangements that move adjudication toward automation.